• No results found

Effects of the Minimum Wage on Employment Dynamics

N/A
N/A
Protected

Academic year: 2021

Share "Effects of the Minimum Wage on Employment Dynamics"

Copied!
53
0
0

Loading.... (view fulltext now)

Full text

(1)

Effects of the Minimum Wage on Employment Dynamics

Jonathan Meer

Texas A&M University

and NBER

Jeremy West

Massachusetts Institute of

Technology

January 2015

Abstract

The voluminous literature on minimum wages offers little consensus on the extent to which a wage floor impacts employment. We argue that the minimum wage will impact employment over time, through changes in growth rather than an immediate drop in relative employment levels. We conduct simulations showing that commonly-used specifications in this literature, especially those that include state-specific time trends, will not accurately capture these effects. Using three separate state panels of administrative employment data, we find that the minimum wage reduces job growth over a period of several years. These effects are most pronounced for younger workers and in industries with a higher proportion of low-wage workers.

Author emails are jmeer@econmail.tamu.edu and westj@mit.edu. We are grateful for valuable

com-ments from Kerwin Charles and two anonymous referees, as well as from David Autor, Jeffrey Brown, Jeffrey Clemens, Jesse Cunha, Jennifer Doleac, David Figlio, Craig Garthwaite, Daniel Hamermesh, Mark Hoekstra, Scott Imberman, Joanna Lahey, Michael Lovenheim, Steven Puller, Harvey S. Rosen, Jared Ru-bin, Juan Carlos Saurez Serrato, Ivan Werning, William Gui Woolston, and seminar participants at the Massachusetts Institute of Technology, the Naval Postgraduate School, Texas A&M University, and the Stata Texas Empirical Microeconomics workshop. We benefited greatly from discussions regarding data with Ronald Davis, Bethany DeSalvo, and Jonathan Fisher at the U.S. Census Bureau, and Jean Roth at NBER. Sarah Armstrong and Kirk Reese provided invaluable research assistance. Any errors are our own.

(2)

1

Introduction

The question of how a minimum wage affects employment remains one of the most widely studied – and most controversial – topics in labor economics, with a corresponding dispute in the political sphere. Neoclassical economic theories present a clear prediction: as the price of labor increases, employers will demand less labor. However, many recent studies testing this prediction have found very small to no effects of the minimum wage on the level of employment (e.g. Zavodny, 2000; Dube et al., 2010; Giuliano, 2013). One possible explanation for these findings is that demand for low-wage labor is fairly inelastic; another is that more complicated dynamics cloud identification of the effect of the minimum wage on employment.1

We argue that there is basis in theory for believing that the minimum wage may not reduce the level of employment in a discrete manner. We show that if this is indeed the case, then traditional approaches used in the literature are prone to misstating its true effects. We also demonstrate that a common practice in this literature – the inclusion of state-specific time trends as a control – will attenuate estimates of how the minimum wage affects the employment level. Specifically, we perform a simulation exercise which shows that if the true effect of the minimum wage is indeed in the growth rate of new employment, then even real causal effects on the level of employment can be attenuated to be statistically indistinguishable from zero.

To implement our analysis, we use a number of different empirical approaches to examine effects of the minimum wage on employment growth and levels; broadly, all of our approaches leverage a difference-in-differences identification strategy using state panels. We perform numerous robustness checks to test the validity of our identification strategy, which requires that the pre-existing time-paths of outcomes for states which increase their minimum wages do not differ relative to states that do not see an increase. We evaluate this possibility by adding leads of the minimum wage into our specifications; if increases in the minimum wage showed a negative effect on employment dynamics before their implementation, this would suggest that the results are being driven by unobserved trends. This is not the case. Indeed, for our results to be driven by confounders, one would have to believe that increases in the minimum wage were systematically correlated with unobserved shocks to that state in the same time period, but not other states in that region, and that these shocks are not reflected in measures of state-specific demographics or business cycles. Our results are

1Hirsch et al.[2011] andSchmitt[2013] focus on other channels of adjustment in response to increases in

(3)

additionally robust to varying the specifications to account for finer spatial and temporal controls, the recent financial crisis, and inflation indexing of state minimum wages, as well as across different panel lengths and time periods.

We use three administrative data sets in our analysis: the Business Dynamics Statis-tics (BDS), the Quarterly Census of Employment and Wages (QCEW), and the Quarterly Workforce Indicators (QWI). These data sets vary in their strengths and weaknesses, dis-cussed at length below, but together they encompass a long (1975-2012) panel of aggregate employment metrics for the population of employers in the United States. Our findings are consistent across all three data sets, indicating that employment declines significantly in response to increases in the minimum wage over the span of several years.

Finally, we find that the effect on job growth is concentrated in lower-wage industries, among younger workers, and among those with lower levels of education. Much of the existing literature focuses on these groups, though it is important to note that the minimum wage could affect other industries or elsewhere in the age and education distributions (e.g. Neumark et al., 2004).

If the minimum wage is to be evaluated alongside alternative policy instruments for in-creasing the standard of living of low-income households, a more conclusive understanding of its effects is necessary. The primary implication of our study is that the minimum wage does affect employment through a particular mechanism. This is important for normative analysis in theoretical models (e.g. Lee and Saez, 2012) and for policymakers weighing the tradeoffs between the increased wage for minimum wage earners and the potential reduction in hiring and employment. Moreover, we reconcile the tension between the expected theoretical effect of the minimum wage and the estimated null effect found by some researchers. We show that because minimum wages reduce employment levels through dynamic effects on employment growth, research designs incorporating state-specific time trends are prone to erroneously estimated null effects on employment. In contrast, the minimum wage significantly reduces job growth, at least in the context that we are able to analyze.

This article proceeds as follows: in Section 2 we provide a brief review of the literature on the employment effects of the minimum wage and build our case for examining employ-ment dynamics. Section 3 presents our econometric models and demonstrates that existing approaches used in this literature obtain incorrect results if the true effect of the minimum wage is on the growth rate of employment. Section 4 describes the data used in our study and presents empirical results. We conclude in Section 5.

(4)

2

Theoretical and Empirical Framework

The economic literature on minimum wages is longstanding and vast. Neumark and Wascher [2008] provide an in-depth review of the field, which continues to be characterized by dis-agreement on how a minimum wage affects employment. The majority of recent studies, following Card and Krueger [1994], use difference-in-differences comparisons to evaluate the effect of these policies on employment levels. Recent papers generally focus on modifying the specification to improve the quality of the counterfactual comparisons, with disagree-ment on appropriate techniques and often-conflicting results (e.g. Allegretto et al., 2011and Neumark et al., 2013). Importantly, these models test whether there is a discrete change in the level of employment before and after a state changes its minimum wage, relative to the counterfactual change as measured by other states’ employment.

Yet there is basis in theory for believing that the minimum wage may not reduce the level of employment in a discrete manner. While the basic analysis of the effects of the minimum wage argues for rapid adjustments to a new equilibrium employment level (e.g. Stigler,1946), transitions to a new employment equilibrium may not be smooth [Hamermesh,1989] or may be relatively slow [Diamond,1981;Acemoglu,2001]. In this case, the effects of the policy may be more evident in net job creation.2 In worker search-and-matching models (e.g. Van den Berg and Ridder,1998; Acemoglu,2001;Flinn,2006, 2011), summarized concisely in Cahuc and Zylberberg [2004], the minimum wage has opposing effects on job creation. Although it reduces demand for labor by raising the marginal cost of employing a new worker, a higher minimum wage increases the gap between the expected returns to employment relative to unemployment, inducing additional search effort from unemployed workers. By increasing the pool of searching workers (and the intensity of their searching), the minimum wage improves the quality of matches between employers and employees, generating surplus. The theory thus has ambiguous predictions for the effect of a minimum wage on job creation. If workers’ additional search effort sufficiently improves the worker-firm match quality, then job creation should not be adversely affected and may even increase. However, if the demand-side effect dominates, then increasing the minimum wage will cause declines in hiring.3

2Of course, any effect on growth does not exclude a discrete effect on the employment level. We separate

these types of effects in the illustrations that follow to facilitate clearer exposition.

3With our reduced-form empirical analysis, we cannot distinguish the true mechanism driving the

rela-tionship between the minimum wage and employment. For instance, it is possible possible that the minimum wage would discretely affect employment, but that frictions in the labor market cause this effect to manifest over time. At a practical and policy-relevant level, these two situations are equivalent, and we are agnostic on the underlying mechanism which, as we discuss in Section4.4, limits our ability to make sweeping statements about how the minimum wage truly impacts labor markets.

(5)

Sorkin [2013] builds a model that formalizes this potentially slow adjustment of labor demand, focusing on firms’ difficulties in adjusting their capital-labor ratios, and applies it to minimum wage increases. He argues that “the ability to adjust labor demand is limited in the short run” and that this “provide[s] an explanation for the small employment effects found in the minimum wage literature.” Fundamentally, this identification problem stems from the “sawtooth pattern” exhibited in states’ real minimum wages. Sorkin argues that “difference-in-difference faces challenges in measuring the treatment effect of interest, which in this case is the effect of a permanent minimum wage increase, whenever there are dynamic responses to the treatment and the treatment itself is time-varying.”

To be clear, if the true effect a minimum wage is to change the slope for employment growth, rather than the employment level, then the traditional approaches used in this literature – namely, difference-in-differences estimates of the effects of the minimum wage on employment levels – will yield incorrect inference.4

2.1

Staggered Treatments and Difference-in-Differences

We illustrate this potential shortcoming of the classic difference-in-differences approach in Figure 1. This toy example depicts employment in two hypothetical jurisdictions, which initially exhibit identical growth rates. At some time t1, Jurisdiction A is treated; at some later time t2, Jurisdiction B is treated with the same intensity. In Panel (a), treatment has a discrete and symmetric negative effect on the employment level, whereas in Panel (b), the treatment has a symmetric negative effect on employment growth, but does not discretely alter the employment level. Consider the standard difference-in-differences (DiD) identification of the employment effect:

Employmentit =δB·I{Jurisdiction = B}+τt·I{Time =t}+β·I(Treatmentit = 1) +uit

Because both jurisdictions are initially untreated and both are eventually treated, the only time period(s) in which the treatment effect β may be identified separately from the time fixed effects τt are those during which only Jurisdiction A is treated. During all other

time periods, I(Treatmentit= 1) takes the same value for both states. Thus, the DiD model

compares the average difference in employment between the jurisdictions during the time

4Several recent studies are exceptions to the focus on employment levels. Dube et al.[2011] examine the

relationship between the minimum wage and employee turnover for teenagers and restaurant workers using the 2001-2008 Quarterly Workforce Indicators (QWI).Brochu and Green[2013] assess firing, quit, and hiring rates in Canadian survey data. Both studies find a reduction in hiring rates but do not estimate the effect on net job growth.

(6)

period betweent1 and t2 to that in the time periods prior to t1 and following t2.

This evaluation is obvious for the discrete employment effect in Panel (a). The difference between jurisdictions’ employment is clearly smaller during the middle time period, compared to the outer time periods, and the DiD estimate is correctly some negative number. Moreover, the duration of each of the three time periods is irrelevant for obtaining the correct inference. If instead the treatment effect is on growth as in Panel (b), then DiD is very sensitive to the relative duration of each (outer) time period. To highlight this sensitivity, consider first the extreme case in which there is a long pre-treatment timespan between times zero and t1, but a very short timespan betweent2 and T, the end of the sample period. In this situation, the average difference in employment during the outer time periods is determined nearly entirely by the pre-treatment period, and the DiD estimate for the treatment effect will be negative. Contrast this with the other extreme: a very short timespan between times zero and t1, but a long period following t2, during which both jurisdictions are treated. In this situation, the average difference in employment during the outer time periods is determined nearly entirely by the later period, and the DiD estimate for the effect of thesametreatment will be positive. And, ifT is selected such that the two outer periods have equivalent duration (i.e. t1−0 =Tt2), then DiD yields a null treatment effect, visibly at odds with the plotted time paths of employment.

This toy example underscores the pitfalls in using a standard difference-in-differences model to identify treatment effects if there is staggered treatment intensity and the treatment affects the growth of the outcome variable. As a state-level policy, the minimum wage clearly exhibits this type of staggered treatment: Figure2 (along with AppendixC) shows that the effective minimum wage changed in at least one state in 33 of the 37 years from 1976 through 2012 – more than 700 changes in total – including every year after 1984.5 We investigate the implications of this concern more thoroughly using Monte Carlo simulation in Section3. First, though, we discuss a separate but related concern.

5Inflation is an additional consideration when evaluating the minimum wage as a policy treatment.

His-torically, minimum wages have been set in nominal dollars, with their value eroding substantially over time (see AppendixCfor details). This means that the actualintensity of treatment changes over time, even in the absence of any subsequent (own or counterfactual) explicit policy change. This situation would not be problematic if the minimum wage affected employment in an abrupt, discrete manner. But if the minimum wage predominantly affects job creation, then it may take years to observe a statistically significant difference in the level of employment. In Section4.4, we revisit the implications of inflation for minimum wage policy in the context of our empirical findings.

(7)

2.2

Implications of Jurisdiction Time Trends as Controls

Many recent studies of the minimum wage include state- or county-specific time trends to control for heterogeneity in the underlying time-paths by which labor markets evolve within different areas that might be correlated with treatment intensity (e.g. Page et al., 2005; Addison et al., 2009; Allegretto et al., 2011). These models generally find little or no effect of the minimum wage on employment levels. However, if the policy change affects the growth rate of the response variable, rather than its level, then specifications including jurisdiction-specific trends will mechanically attenuate estimates of the policy’s effect. The basic intuition is that including state-specific time trends as controls will adjust for two sources of variation. First, if there is any pre-treatment deviation in outcomes that is correlated with treatment – e.g. if states that exhibit stronger employment growth are also more likely to increase their minimum wage – then this confounding variation may be appropriately controlled for by including state-specific time trends. The potential cost of this added control is that if the actual treatment effect, thepost-treatment employment variation, acts upon the trend itself, then inclusion of jurisdiction time trends will attenuate estimates of the treatment effect and often leads to estimating (statistical) null employment effects.6

A simple illustration of this is provided in Figures 3and 4. Figure 3depicts employment in two hypothetical jurisdictions which exhibit identical employment growth rates prior to period t = 0 . After period t = 0, the employment growth rate in the Treated jurisdiction falls relative to the Control, but there is no discrete change in the level of employment. Figure 4 presents the difference in employment by time period for both levels and growth, with and without adjustment for jurisdiction time trends. The computed employment effect is large and negative when state trends are omitted (in Panel (a)), but shrinks nearly to zero with the inclusion of jurisdiction time trends (Panel (b)). This occurs despite identical pre-treatment employment trends. In contrast, inference about the effect on employment growth is the same regardless of whether the the data are detrended (Panels (c) and (d)), because the effect on growth is discrete.

We are by no means the first to make this point. In examining the effects of changes in divorce laws, Wolfers [2006] makes a general observation that a “a major difficulty in difference-in-differences analyses involves separating out trends from the dynamic effects of a policy shock.” Lee and Solon[2011] expound on this point in a discussion ofWolfers[2006], pointing out that “the sharpness of the identification strategy suffers” when jurisdiction-specific time trends are included and, “the shift in the dependent variable may vary with the

(8)

length of time since the policy change.” This problem has been discussed in other contexts, including bias in estimates of the effects of desegregation (Baum-Snow and Lutz, 2011) and marijuana decriminalization (Williams,2014).

However, this approach remains common in the minimum wage literature and, indeed, for many other important policy questions in which researchers ask “a much more nuanced question than just whether the dependent variable series showed a constant discrete shift at the moment of policy adoption” (Lee and Solon, 2011). We hope that our examples and simulations will serve as a useful guide to researchers considering how to approach estimation of policies whose effects may differ over time and, especially, may be reflected in changes in the growth rate of the variable of interest. We delve further into the question of how best to estimate these effects in Section 3.

3

Econometric Specifications and Simulations

In Section 2, we provide theoretical support for the hypothesis that the minimum wage affects the growth rate of employment, even if it does not induce a discrete drop in the level of employment, and we illustrate several complications for attempts to empirically quantify the magnitude of such an employment effect. In this section, we present several econometric models as candidates to estimate this effect, comparing their strengths and shortcomings both analytically and using simulated data in a Monte Carlo framework. The goal of this section is not to argue for one “correct” model to estimate the relationship between the minimum wage and employment, but rather to underscore the tradeoff between the various assumptions that can be invoked in order to obtain causal inference about this treatment effect.

3.1

Candidate Specifications

Consider the following panel difference-in-differences model relating the minimum wage to employment:

empit =αi+τt+γi·t + s

X

r=0

βrmwit−r+ψ·controlsit+it

in whichempit is the level of employment in stateiat timet,αi are jurisdiction fixed effects,

(9)

trends, and it is the idiosyncratic error term.

If the true treatment effect is fully discrete in levels, as in the scenario depicted in Panel (a) of Figure 1, then βr = 0 ∀r > 0, as lags of the minimum wage do not separately affect

the current employment level. The model reduces to:

empit =αi+τt+γi·t +β0mwit+ψ·controlsit+it (1)

and the estimate ˆβ0 identifies the total causal impact of the minimum wage on employment. Specification 1 is the “classic” variant of the difference-in-differences specification, in levels, and has been used extensively in the literature.

In contrast, if the true treatment effect instead acts on the growth rate of employment, as in the scenario depicted in Panel (b) of Figure 1, then βr 6= 0 for at least some lagged

values of the minimum wage. The full set of lag terms are necessary, yielding a distributed lag model in levels:

empit =αi+τt+γi·t + s

X

r=0

βrmwit−r+ψ·controlsit+it (2)

An alternate approach is to difference the model, yielding the distributed lag model in first-differences:

∆empit=θt+γi + s

X

r=0

βr∆mwit−r+ψ·∆controlsit+ ∆it (3)

Either Specification 2 or Specification3 can be used to flexibly identify the dynamics of the effect of the minimum wage on employment, and summing the βr identifies the overall

effect on the employment level. Whether it is preferable to estimate distributed lag coeffi-cients using a fixed effects versus a first-differenced model is not clear.7 Nichols [2009] notes that a major consideration in this decision is the timing between the change in treatment and the observed effect, the theoretical relationship of which is not obvious in this context. Moreover, depending on the degree of serial correlation betweenit and between ∆it, either

Specification 2 or 3 may be more efficient; as Wooldridge [2002] notes, the “truth is likely to lie somewhere in between.” Our focus on importance of changes from year to year, as opposed to comparing differences in pre- and post-periods, suggests that the first-differenced approach is more appropriate in this case. Nevertheless, we leverage both variations of the

7An additional consideration is that the asymptotic properties of the fixed effects estimator rely on

(10)

distributed lag model, testing them in the Monte Carlo simulation below and presenting both in the primary results tables.

Although distributed lag models such as these are relatively common in the program evaluation literature, both forms of the specification suffer from a common shortcoming when examining minimum wage effects. Specifically, the high frequency variation in treatment intensity makes it difficult to make credible causal inference about the employment effects of higher-order lags of the minimum wage, because the large number of changes and potential long-run confounders make a fully-specified model fragile. Put another way: in practice the number of included lags s must be fairly small in any distributed lag specification, in either levels or first-differences. Including only a short number of lag terms reduces the utility of using a distributed lag specification to estimate an effect on growth.

Given this restriction on the number of lag terms that can sensibly be included, a natural approach is to use a dynamic panel specification (e.g. Arellano and Bond, 1991). This allows us to estimate both the short- and long-run effects, at the cost of imposing a stricter assumption on the nature of this relationship. The specification then takes the form:

empit =µ·empit−1+αi+τt+γi·t + s

X

r=0

βrmwit−r+ψ·controlsit+it

which differs from the above models in that the lag of employment is included on the right hand side. This can be first-differenced to eliminate theαi jurisdiction fixed effects:

∆empit=µ·∆empit−1+θt+γi + s

X

r=0

βr∆mwit−r+ψ·∆controlsit+ ∆it (4)

In this dynamic panel model, the short run marginal effect of the minimum wage on employ-ment isβ0, and the effect after one year of a sustained change is captured byβ1+ (1 +µ)∗β0. Due to the properties of a geometric series, the long run effect on employment is determined by (β0 +β1)/(1−µ). Importantly, this long run effect (in fact, the specific time path of the effect) can be identified using only a single lag term for the minimum wage. Thus, a dynamic panel specification skirts much – although not all – of the concern about constantly changing treatment intensities.8

8In solving one identification problem, the dynamic panel approach introduces another, as the ∆emp

terms are autocorrelated. The standard practice, as in Arellano and Bond [1991], is to use deeper lags of employment as instruments for the lagged employment term. However, these may not be exogenous, depending on the degree of autocorrelation. As we discuss later, our results are robust to a number of approaches, including the use of deeper lags of theminimum wage rather than employment as instruments. We are grateful to an anonymous referee for suggesting this approach.

(11)

We have yet to discuss the role of the jurisdiction time trends, γi·t, in comparing these

specifications. Provided the true treatment effect is fully discrete in levels, then including jurisdiction time trends will not bias the estimated ˆβ0 in any of the above models (recall that for an effect that is fully discrete in levels, βr = 0 ∀r > 0). Jurisdiction-specific time

trends can be included as controls for any underlying variation in employment trends – which might be correlated with treatment intensity – without biasing the estimate for the

β0 parameter of interest. However, if the true treatment effect instead acts on the growth rate of employment, then including jurisdiction time trends will bias estimates in all of the above models. In this case, because the minimum wage actually affects the slope of the employment trend, including jurisdiction-specific time trends in the specification will directly bias estimates of the βr parameters of interest.9

One possibility to avoid this bias would be to identify the jurisdiction-specific time trends using only pre-treatment time periods: that is, to estimate γi for each jurisdiction during

the pre-treatment period only, and then extrapolate these trajectories throughout the entire study timeframe. This approach may work well for many studies in the program evaluation literature, in which treatments are usually discrete one-time changes. However, the validity of this approach requires that there actually is a sufficient pre-treatment period, a condition that demonstrably fails to hold in the case of the minimum wage in the United States. In this context, this first option is off the table.

A second option is to test for the presence of pre-treatment variation in employment trends directly by using a common “leading values” falsification test, and – provided this test is passed – simply exclude jurisdiction-specific time trends from the specification. Recall that the concern is that jurisdictions which disproportionately increase their minimum wage might have had comparatively negative employment trends even in the absence of differences in treatment. If the econometric test reveals that this is unlikely to be the case, then the model can be changed to forceγi = 0 ∀i. Theβr terms will yield unbiased estimates of the

distributed lag effects of the minimum wage provided that jurisdiction time trends are not of importance in the true model.

Testing for pre-treatment deviation in outcomes should alleviate concerns about the im-portance of controlling for heterogeneity in jurisdiction time trends. But, if it remains unpalatable to eliminate jurisdiction time trends entirely from the model (and provided the treatment effect is on growth), then the remaining option is to impose an additional strong

9Note that jurisdiction time trends would still bias estimates for a treatment effect on growth even if it

were possible to fully saturate the model with post-treatment lags of the treatment variable. The fundamental issue stems from the treatment affecting the trend itself, as illustrated earlier in Figures3 and4.

(12)

restriction by setting β0 ≡β1 ≡ ...βs. This restriction requires that the minimum wage

affect employmentgrowth discretely and permanently – that there is not a dynamic relation-ship between the minimum wage and employment growth. This restriction is consistent with the relationship depicted in Figure 3, in which the minimum wage causes a break-in-trend for the employment level, rather than a discrete drop in the employment level. Provided that this assumption holds, then:

s

X

r=0

βr∆mwit−r =β0·(∆mwit+ ∆mwit−1 +...+ ∆mwit−s) = β0·mwit

and Specification 3 is equivalent to:

∆empit=θt+γi +β0·mwit+ψ·∆controlsit+ ∆it (5)

Specification 5, which we refer to as the “break-in-trend” model, is the only specification of these five that is robust to including jurisdiction time trends without biasing estimates ofβr

for a treatment effect on growth. This distinction comes at the cost of a strong assumption about the nature of the dynamics of the treatment effect. In practice, it seems very unlikely that the minimum wage wouldpermanently reduce the growth rate of employment – indeed, extrapolating such an effect far into the future would predict immense employment effects. For this reason, we primarily view Specification 5 as a trends-robust indication of whether

the minimum wage affects the growth rate of employment, with the possibility of calcu-lating back-of-the-envelope estimates of the magnitudes of proposed policy, given certain assumptions. We return to this issue in detail in Section 4.4.

We will present results from each of these five specifications – classic, distributed lags in levels, distributed lags in first-differences, dynamic panel, and break-in-trend – both with and without including jurisdiction time trends, for all three data sets, in Section 4. First, though, we use Monte Carlo repetitions of a fairly simple simulation to underscore how severely time trends bias estimates of an effect on growth across these specifications.

3.2

Monte Carlo Simulation

In this section, we conduct a Monte Carlo exercise with simulated data to compare the efficacy of the five models and to illustrate how severely including jurisdiction time trends biases estimates when the treatment effect is on growth.

(13)

and employment (in the Business Dynamics Statistics data, discussed below in Section 4). Drawing without replacement from these data, we form two independent distributions of changes, one for real minimum wages and one for employment. We merge these distributions together to form a new panel containing 35 periods for 51 state entities, repeating this process within each Monte Carlo repetition.

Next, we impose a treatment effect relating the minimum wage to the growth rate of employment. To prevent the effect from being purely deterministic, we draw the treatment effect from a N ormal(−0.03,0.015) distribution for each state-year observation. That is, each 10% increase in a state’s real minimum wage causes, in expectation, a 0.3 percentage point reduction in employment growth. Because the effect is on the employment growth rate, the treatment effect in a state in one year persists throughout all future years, a pattern such as that illustrated earlier in Figure 3. While imposing this type of treatment effect is extreme – an increase in the real minimum wage will permanently reduce the growth rate of employment – it facilitates clarity in comparing the five models and highlighting the concern with jurisdiction time trends.

With these simulated data, we estimate the relationship between the minimum wage and employment using each of the five specifications, separately with and without including jurisdiction time trends. Table 1 reports the median coefficients from 10,000 Monte Carlo repetitions of these estimations.10 Consider first Column (1) in Panel [A], which excludes jurisdiction time trends. The standard difference-in-differences model clearly identifies a negative average treatment effect, though this coefficient does not clarify whether the treat-ment discretely affects the level of the outcome or if it affects the growth rate. In contrast, when time trends are added in Panel [B], the coefficient in Column (1) is attenuated to, essentially, a zero estimated treatment effect. This occurs despite the fact that time trends cannot actually be helpful for these estimations, because the simulated data have random employment shocks that are by construction only correlated with minimum wages through the imposed treatment effect.

The dynamics of the treatment effect are more salient in the distributed lag specifications in Columns (2) and (3) of Panel [A]: it is clear that the treatment does not simply induce a one-time contemporaneous drop in the level of the outcome, but instead continues to negatively affect employment in future periods, i.e. an effect on growth. The pattern in Column (2) for the estimated dynamics when using distributed lags in levels shows that

10The full code used in this simulation, along with all other code and data included in this study, is

(14)

there is no contemporaneous effect on the employment level and an increasing cumulative effect over time, with the final lag term capturing the remaining average treatment effect. We somewhat arbitrarily opted to include only three lag terms, but this basic pattern of a “zero” contemporaneous effect and a large final lag term holds regardless of the number of lag terms included in Specification 2, be it one or many. The important thing to note is that this approach does not yield accurate results, either, though it does highlight that there is a dynamic response following the simulated treatment. Panel [B], which includes time trends, shows an even larger deviation from the true effect, with a relatively large and positive contemporaneous coefficient.

Turning to Column (3), the distributed lags with first differences model accurately cap-tures the constant treatment effect that was imposed on growth. Yet as with the previous two specifications, this model also exhibits attenuation of the estimates when jurisdiction time trends are included in Panel [B].

Column (4) shows the results of the dynamic panel simulation. The autoregressive term for the lag of employment is 0.852, with the contemporaneous minimum wage coefficient equaling -0.019 and the first lag equaling -0.041. This implies that a permanent, real increase in the minimum wage results in a short-run elasticity of -0.019 in the first year and -0.076 in the second year. The long-run effect, calculated as explained above, is -0.407. While this model does require stricter assumptions, the primary advantage is the ability to examine the short- and long-run elasticities; essentially, these results allow us to plot out the effect on the level of employment, showing an initial dip to a new employment level that subsequently runs parallel to that of the counterfactual.11 Much like the previous specifications, including jurisdiction-specific trends in Panel [B] substantially biases the estimates: the short-run impact changes to a positive 0.014 in the first year and 0.019 in the second year – that is, not even the sign is correct – with a very small permanent effect of -0.016.

Finally, the Break-in-Trend specification in Column (5) identifies the nature of the “kink” in the employment time path. We stress again that the accuracy of the estimatedmagnitude

of this coefficient depends on the validity of the strong identifying assumption about a permanent effect on growth (which happens to be true by construction in this simulation). The value of this specification is that – in only this model – the coefficient is not biased when jurisdiction time trends are included, as we showed analytically above and as is evidenced by comparing Panel [A] to Panel [B] of Column (5) in Table 1.

11Note that, in the case of the extreme data-generating process that we impose, this prediction is incorrect.

We discuss the general difficulties of making inference about permanent changes in the minimum wage, especially without imposing model-based restrictions, in Section4.4

(15)

Summarizing the findings of this Monte Carlo simulation, we have shown that – if the true treatment effect is on employment growth – including jurisdiction time trends can starkly bias estimates from a difference-in-differences specification, whether the model is the classic form, a distributed lag specification, or even a dynamic panel model. For exposition, we simulated the extreme case of a permanent treatment effect of the real minimum wage on employment growth. However, our findings generalize to drawing entire minimum-wage histories rather than individual-year changes; to allowing the minimum wage treatment intensity to be eroded due to inflation; to introducing underlying jurisdiction-specific trends that are correlated with whether the jurisdiction has a high or low minimum wage; and to treatment effects that attenuate over time. Most importantly, this simulation exercise contrasts the various specifications that we will estimate in the next section and illustrates the general pattern of results to be expected of an effect on employment growth.

4

Empirical Results

4.1

Data

We estimate employment effects using three data sets: the Business Dynamics Statistics

(BDS) and theQuarterly Workforce Indicators (QWI), both from the Bureau of the Census, and the Quarterly Census of Employment and Wages (QCEW) from the Bureau of Labor Statistics. The QCEW and QWI report quarterly employment for each state, while the BDS is annual. All of these data are administrative in nature; the QCEW and QWI programs collect data from county unemployment insurance commissions, while the BDS reports on employment rosters furnished to the U.S. Internal Revenue Service. As such, each of the data sets we study accounts for virtually the entire population of non-farm employment.12 For brevity and clarity of exposition, we report results from the BDS in the main body of the paper, with results from the full set of specifications using the QCEW and QWI in Appendix

12The employer-sourced administrative nature of these data is important for our research question.

Population-level data provide for a cleaner assessment of the overall policy impact of minimum wages by avoiding sampling error. Moreover, as discussed in Section2, a higher minimum wage may induce additional searching effort on the part of the currently unemployed. Mincer [1976] shows that this positive supply elasticity often leads to an increase in the number ofunemployed that differs substantially from the change in employment. Because employment is the policy-relevant outcome, measuring job counts using employer-sourced data provides a better identification of any disemployment effects than do surveys of individuals, such as the Current Population Survey. Finally, employment data directly reported by firms to maintain legal compliance have been shown to be more accurate than responses to individual-level surveys such as the CPS [Abraham et al.,2009].

(16)

A. As we note below, there is little difference in the overall results across the three data sets, which is unsurprising given that all three examine the near-population of jobs in the United States.

The BDS covers all non-agriculture private employer businesses in the U.S. that report payroll or income taxes to the IRS. The heart of the BDS is the Census Bureau’s internal Business Register, which is sourced from mandatory employer tax filings and augmented using the Economic Census and other data to compile annual linked establishment-level snapshots of employment statistics (on March 12th). The Census Bureau releases the BDS as a state-year panel (all fifty states, plus the District of Columbia), currently covering 1977 to 2011. Summary statistics from the BDS are provided in Table2. Full descriptions of the QCEW and QWI, including their summary statistics, are located in Appendix A.

4.1.1 State Minimum Wages

We draw historical data on state minimum wages from state-level sources.13 For the QCEW and QWI, we use the minimum wage value as of the first of each quarter. For the BDS, we use the value as of the previous March 12th each year, directly corresponding to the panel years in the BDS data. Some states have used a multiple-track minimum wage system, with a menu of wages that differ within a year across firms of different sizes or industries; we therefore use the maximum of the federal minimum wage and the set of possible state minimum wages for the year. To the extent that there is firm-level heterogeneity in the applicable wage level, our definition allows the minimum wage term to serve as an upper bound for the minimum wage a firm would actually face. We transform minimum wages into constant 2011 dollars using the (monthly) CPI-U from the Bureau of Labor Statistics.14

4.1.2 Other Control Variables

Although our econometric specifications include an extensive set of time period controls, precision may be gained by accounting for additional state-specific time-varying covariates.

13Although historical state minimum wage data are available from sources such as the U.S Department of

Labor (http://www.dol.gov/whd/state/stateMinWageHis.htm), these data suffer several limitations. For one, minimum wage values are only reported only as of January first each year, whereas the panel used in our study necessitates values as of other dates. Additionally these DOL data incompletely characterize changes to state minimum wages, especially during the early years of our panel. This DOL table is frequently used as the source of historical state minimum wage values for recent studies in this literature, and we caution future researchers to be careful not to inadvertently attribute minimum wage changes to years in which they did not occur.

14Because we use a national-level deflator, specifying the log minimum wage term as real or nominal does

(17)

The Census Bureau’s Population Distribution Branch provides annual state-level population counts, including estimates for intercensal values. Total state population represents a deter-minant of both demand for (indirectly by way of demand for goods and services) and supply of employees. Because states differ non-linearly in their population changes, controlling di-rectly for population may be important. The range in population between states and across time is enormous, so we use the natural log of state population in our specifications. We additionally include the share of this population aged 15-59, which provides a rough weight for how population might affect demand for versus supply of labor. Demographic controls such as these are commonly used in this literature (e.g. Burkhauser et al.,2000; Dube et al., 2010). Following Orrenius and Zavodny [2008], we also include the natural log of real gross state product per capita.15 After controlling for state population, this term can be thought of as a rough proxy for average employee productivity as well as a measure of state-level fluctuations in business cycles [Carlino and Voith, 1992, Orrenius and Zavodny, 2008].

4.2

Results

We begin with a very simple diagnostic check: if the true effect of the minimum wage is on growth, then specifications that are differenced over increasingly long time periods should yield larger coefficients for the effect of the minimum wage on employment. We take Equation 3with a single minimum wage term, and increase the number of years over which we difference the equation. Indeed, this simple check shows evidence for effects on growth: the coefficient on the minimum wage term for a one-year difference is -0.020 (s.e. = 0.018); taking the difference from two years previously changes the coefficient to -0.039 (s.e. = 0.021); for three years, it is -0.050 (s.e. = 0.024); for four years, it is -0.051 (s.e. = 0.024). The coefficient is stable around this magnitude even when differencing by as much as eight years, and similar results are seen in the QCEW and QWI. While this diagnostic does not provide definitive proof that the effects of the minimum wage are on growth – after all, many other factors can change over such long periods – the absence of such a pattern could be taken as evidence against our hypothesis.

In Table 3, we present results for the five specifications from Section 3 to identify the effect of the minimum wage on employment using the Business Dynamics Statistics (as mentioned above, results for the QCEW and QWI are available in Appendix A). Of course,

15We compute the log of the real value of total GSP per capita using all industry codes, including

gov-ernment. Results are virtually unaffected by using ln(realprivate sector GSP/capita) instead, but we view total GSP as the more appropriate definition given that the population term reflects total state population.

(18)

estimations using the actual data do not generate coefficients that are as tidy as those using a prescribed data-generating process.16 Nevertheless, the models in Section3that are shown to accurately capture effects on growth yield similar estimates in all three data sets, and, broadly, estimates across all specifications show similar patterns to their counterparts using the artificially-generated data.

We focus first on Panel [A], which excludes the jurisdiction-specific trend terms. The classic difference-in-differences model in Column (1), which corresponds to Specification 1 in Section 3, shows a significant disemployment effect of the minimum wage. Specifically, the estimate is that a permanent ten percent increase in the real minimum wage causes about a 1.7 percent decline in total employment. As with the results from the Monte Carlo simulation, the classic model cannot distinguish between an effect on growth and a discrete effect on the employment level. The dynamics of the treatment effect are more apparent in the distributed lag model in Column (2). It is clear that the effect is not encompassed in a one-time discrete drop in employment; rather, the minimum wage appears to have a fairly constant negative effect on the growth rate of employment over the period covered by the lags. The effects for each minimum wage coefficient are negative and, with the exception of the third lag, statistically significant. A permanent increase in the minimum wage, according to this model, would yield an employment elasticity of -0.29 (s.e. = 0.06). In Column (3), the distributed lag model in first differences, we see a fairly steady and negative impact of the minimum wage on employment, similar to the one found in Table1; the third lag is positive and statistically insignificant, suggesting that the effects of a minimum wage change fade out after about three years, though this pattern could also result from the high-frequency variation in minimum wage changes. Importantly, much of the impact comes in the two yearsafter the change, suggesting that short-term data immediately after an increase in the minimum wage is unlikely to show its true impact. Summing up these coefficients yields the effect of a permanent change: -0.074 (s.e. = 0.036).17 Irrespective of the magnitudes, we view the results in these two columns as strong evidence that the effect of the minimum wage on employment is of a more dynamic nature than that supposed in the frictionless

16This reduced precision is partly a (lack of) Law of Large Numbers issue: the simulation had 10,000

repetitions of 1785 observations to obtain those coefficients, whereas these results have only the 1785 real-world observations, based on 51 jurisdictions . In addition, unlike in the simulation, real minimum wages are not randomly assigned: there is strong bunching of changes around certain years, for instance. Finally, the simulation prescribed a simple effect just on employment growth, whereas the minimum wage in practice could affect both the level and growth of employment.

17Additional lags do not make a qualitative difference to the sum of coefficients, and the coefficients on

(19)

neoclassical framework. This is further evidenced by the dynamic panel specification in Column (4).18 The contemporaneous elasticity of a minimum wage increase is -0.031 (s.e. = 0.017), with the lag term (-0.054, s.e. = 0.02) implying that the impact after one year at the same treatment intensity would be -0.10 (s.e. = 0.033) and after two years, -0.14 (s.e. = 0.49); the long-run impact of a permanent real increase in the minimum wage effect is -0.20 (s.e. = 0.088).

Contrast these results with those in Panel [B], in which jurisdiction-specific trend terms are included. Across the first four models, the coefficients are sharply attenuated and few remain statistically different from zero. Given the clear evidence in Panel [A] that the effect is not discrete on the employment level, this attenuation is exactly what we would expect based on the theoretical and econometric arguments made in Sections 2 and 3. Moreover, the pattern to this contrast between Panels [A] and [B] of Table3closely mirrors that shown in the Monte Carlo simulation in Table 1: including jurisdiction trends mechanically biases the estimated coefficients across all four models.

Finally, consider the Break-in-Trend model in Column (5). Note that to ensure identifi-cation is coming from within-jurisdiction changes in minimum wage, we include the initial minimum wage by jurisdiction as an additional control in Panel [A].19The strong assumptions underlying this specification require caution in drawing causal inference about themagnitude

of the estimated employment effect.20 That said, it is reassuring that this model yields an estimate in Panel [A] that is similar in magnitude to the per-period coefficients identified for

18The standard approach is to use deeper lags of the dependent variable as instruments [Arellano and

Bond,1991]. Concerns about endogeneity suggest using deeper lags of the minimum wage values themselves as instruments. We use Roodman’s (2009) Stata module, which allows for flexible estimation of dynamic panel models, using this approach, though the coefficients on the minimum wage terms are stable across different sets of instruments. We are grateful to an anonymous referee for this suggestion.

19The essence of the difference-in-differences identification strategy is to identify the effect using temporal

variation within jurisdiction, rather than between jurisdictions. Whereas Columns (1)-(4) either include a jurisdiction fixed effect or first-difference the minimum wage term, Specification (5) in Panel [A] does neither. In the absence of a jurisdiction fixed effect (which is added to Specification (5) in Panel [B]), including the initial minimum wage by jurisdiction controls for heterogeneity in the baseline differences in jurisdictions’ minimum wages and ensures that identification comes from within-jurisdiction variation. This was not an issue in the simulation, for which initial minimum wage values were randomly assigned.

20This is not to say that the results do not hold implications for nominally-set minimum wages. One

reasonable approach is to apply the average “erosion” rates of the minimum wage in the data (see Appendix

C for historical minimum wage erosion rates, as well as the discussion in Section 4.4). Suppose that a state increases its nominal minimum wage by 10% relative to other states within its Census region. The average erosion rate in our panel predicts a remaining effective difference of 6.64% after one year. This relative difference shrinks to 3.87% by the next year, to 2.31% the year after, and to 0.84% after four years, before fully eroding. This suggests a cumulative that is 2.37 times the coefficient in the break-in-trend graph, implying a long-run employment elasticity for the type of minimum wage increases seen in the data of -0.064.

(20)

first-differences in the distributed lag model in Column (3). Perhaps more importantly, the estimated coefficient in Specification (5) changes little (and remains statistically equivalent) when jurisdiction trends are included in Panel [B]. We view this as further evidence that trends are not a confounding factor, but if anything, the slight increase in magnitude shows that estimates are biased towards zero when trends are omitted from these models.

4.3

Additional Specifications and Robustness Checks

In this section, we present a number of alternative specifications to assess the robustness of our empirical results. Most importantly, we perform the common leading-values falsification test for pre-treatment deviation in employment outcomes, thereby examining the validity of the key identifying assumption underlying the difference-in-difference methodology.21 In addition, we show that our results are consistent for different time periods within our sam-ple, and demonstrate invariance of our results to allowing for finer spatial and time controls, accounting for minimum wage inflation indexing, and dropping the years of the recent fi-nancial crisis. For these additional results, we present estimates using Specification 3, the distributed lag model in first-differences, and Specification 5, the Break-in-Trend model, as these two specifications most accurately identify the effect on growth in the Monte Carlo simulation.

Robustness checks using Specification3 are in Table4. Column (1) replicates the results from Column (3) of Table 3: Panel [A], for comparison. Columns (2)-(4) include either the first or second leading value of the minimum wage, or both. If increases in the minimum wage appear to have an effect on employment dynamicsbeforetheir implementation – especially if contemporaneous changes lose their effect – then our results might be driven by unobserved trends. This is not the case: although some precision is lost, the contemporaneous and lagged minimum wage coefficients in Columns (2)-(4) remain close to those in Column (1), and the leading value terms are comparatively small and statistically insignificant. This strongly suggests that confounding trends leading to both lower job growth and higher minimum wages are not a factor. In Column (5), we allow the time effects to vary by Census Division, rather than Region; the coefficients remain similar to those in Column (1). Some precision is lost, though this to be expected – there are four Census Regions containing

21An additional approach to examine the potential endogeneity of minimum wage changes is to examine

the results with different combinations of the time-varying covariates. Results from different combinations of time fixed effects (national versus Region versus Division) and other time-varying controls are stable in magnitude, sign, and significance, particularly across the specifications shown above to accurately reflect minimum wage effects, namely, distributed lags with first differences, dynamic panel, and break-in-trend.

(21)

nine Divisions, and the median Division includes only five states. In Column (6), we assess whether states that have shifted to indexing their minimum wage for inflation affect our results by dropping these observations. Results remain similarly unchanged. Finally, in Column (7), we evaluate the role of the 2008-2009 recession. Because we include time period fixed effects, the recent recession should not unduly affect our results. However, these two years of our panel additionally experienced several large and high frequency changes in real minimum wage levels, primarily resulting from the federal increases during these years (see Figure2). As a check that these particular years are not overly influencing identification of the minimum wage term, we estimate specifications using only pre-2008 data. Again, the estimated effects are not meaningfully different from our main results, though the sum of the minimum wage terms is significant only at p = 0.13; this is somewhat unsurprising given that about fifteen percent of the observations are lost.

Table 5 presents the additional results for the Break in Trend model, Specification 5. Column (1) reproduces the main estimates from Table 3. In Column (2), we include an indicator which equals one if the nominal minimum wage changes the following period. In Columns (3)-(4), we include the leading value of the log of the minimum wage either two or three periods in advance.22 Columns (5)-(7) present, respectively, results using Division-by-year fixed effects, observations without inflation-indexed minimum wages, and pre-2008 data only. As with the distributed lags of first-differences model, these alternative results reflect those in the baseline specification, and the break-in-trend model consistently indicates a statistically significant and economically meaningful effect of the minimum wage on em-ployment growth. Coefficients for the leading indicator or values of the minimum wage again support the validity of the difference-in-differences identifying assumption in this context.

We additionally evaluate the sensitivity of the results to the time period used. For difference-in-differences estimates, there is nearly always a concern that results could be particular to the time-span included in the study. Generally, it seems most appropriate to use all available periods within a data set unless given a compelling reason to do otherwise.

22Note that we do not include a one-period leading value nor include multiple leads simultaneously. This

is because there is explicit collinearity between the current and the lag of the minimum wage term. For simplicity, suppose that the true data-generating process is Yt = β1ln(M Wt) +t. Ordinarily, including

ln(M Wt+1) would show no effect in this regression. However, since Yt= (emptempt−1) and ln(M Wt+1)

is related toYt+1, which includesempt, adding a single-period lead introduces substantial endogeneity. This

is not an issue for leads of at least two periods difference from each other. If the pre-trend identification assumption is violated, it is difficult to believe that it would not be apparent two periods prior as well. Moreover, including a binary variable for whether there is a change in the following period (as opposed to the actual minimum wage value) yields little indication that there is some negative shock that is correlated with both increases in the minimum wage and reductions in job growth.

(22)

However, such an approach cannot guarantee that estimated effects are not particular to the time period used. We evaluate results obtained from estimating Specification 5 separately for all possible subsample spans of two or more consecutive years in the BDS (1977-2011), yielding 595 point estimates using the BDS. We also examine the QCEW, which has 704 such periods.23 Appendix B includes histograms of these coefficients. Sorted by magnitude, the median coefficient is -0.0322 in the BDS, and the first point estimate with a positive value is at the 96th percentile. This exercise indicates that the result of a negative job growth effect of the minimum wage is not simply an artifact of the time spans of data used in this study. We also examine the effects of the minimum wage by industry, age group, and education level in Section A.2.2.

4.4

Discussion

Our results show that the minimum wage negatively affects employment and that this oc-curs over a period of several years. The results from the distributed lag specification in first differences suggest that a 10% permanent increase in the real minimum wage reduces employment by about 0.7 percent after three years. In the dynamic panel model, we leverage additional assumptions to estimate an employment elasticity of about -0.17 after three years and -0.20 in the long run. Taken at face value, our most restrictive model, the break-in-trend specification, suggests that a 10% permanent increase in the real minimum wage reduces job growth by about 0.3 percentage points annually, or about 15 percent of the baseline level. This effect is not small, and extrapolated sufficiently into the future this implies a deleterious effect on employment of enormous magnitude, far surpassing that of any historical recession. The purpose of this section is to caveat our findings and to place the results into perspective for considering the short- and long-run impact of minimum wage policy.

First, we study employment effects in the context of policies in the United States over the past few decades, during which increases in minimum wages have been relatively small in magnitude, albeit frequent. Extrapolating the effect we estimate to the distant future or to much larger increases in the minimum wage requires strong assumptions and is a wildly out-of-sample prediction, one that we refrain from making. Essentially, there is no way – without model-based assumptions – to gain a full understanding of the dynamic responses to a large, real and permanent increase in the minimum wage, because no such change has

23For an initial year of 1977, the BDS has 34 possible spans of at least two years: 1977-1978, 1977-1979,

1977-1980, ..., 1977-2011. An initial year of 1978 has 33 possible such spans, etc. For the QCEW, we could instead consider spans of quarters, but this would not add much in the way of inference. Note that the QWI is too short and too unbalanced to benefit from this exercise.

(23)

ever occurred in the data. This issue is not specific to our study, but does hamper the ability of researchers in this field to make definitive statements on the effects of these policies.

Second, our specifications estimate the relationship between the real minimum wage and employment. Historically, most minimum wage changes have been set in nominal terms and not adjusted for inflation. As we show in AppendixC, inflation substantially erodes the “bite” of a wage floor over time; this is not because nominal minimum wagesaffect employment less significantly than do real minimum wages but rather because the intensity of the policy itself is mitigated. This is illustrated in our brief discussion in Footnote 20, in which we apply the “erosion” rate of the treatment intensity to our break-in-trend specification to evaluate the effect of a nominal minimum wage increase.

The upshot of this distinction is that nominally-denoted minimum wages should have a smaller employment impact than a wage floor that is indexed for inflation. To date, little is known empirically about how inflation indexing may alter the effects of a minimum wage on employment even as at least ten states now use regional CPI measures to index their minimum wages for inflation, a relatively recent practice [Allegretto et al., 2011]. Ongoing minimum wage proposals, such as the federal minimum wage increase proposed by President Obama in 2013, continue to include provisions indexing the wage floor for inflation. As such, this line of inquiry is likely to grow in importance.

5

Conclusion

We examine whether the minimum wage impacts employment through a discrete change in its level or if it is reflected over time through a change in the growth rate. Much of the previous literature on the topic has assumed that an increase in the minimum wage results in a relatively rapid adjustment in employment. Yet, there are theoretical reasons to believe that this change may be slower. Using both illustrative models and Monte Carlo simulations, we show that the empirical specifications used in the prior literature will systematically err if the true effects are on growth rates. Moreover, we show that the common practice in this literature of including jurisdiction-specific time trends will bias estimates towards zero in this case.

We show results from three administrative data sets that consistently indicate negative effects of the minimum wage on job growth. Our results are robust to a number of spec-ifications, and we find that the minimum wage reduces employment over a longer period of time than has been previously examined in the literature. This phenomenon is

(24)

particu-larly important given the evidence that minimum wage jobs often result in relatively rapid transitions to higher-paying jobs [Even and Macpherson, 2003].

This paper, of course, does not settle the debate of a contentious topic, but we do shed light on the mechanisms by which the minimum wage affects employment and provide directions for future research delving more deeply into the dynamics of this relationship.

References

John M. Abowd and Lars Vilhuber. Statistics of jobs. Mimeo, February 2013.

Katharine G. Abraham, John C. Haltiwanger, Kristin Sandusky, and James Spletzer. Ex-ploring differences in employment between household and establishment data. Working Paper 14805, National Bureau of Economic Research, March 2009.

Daron Acemoglu. Good jobs versus bad jobs. Journal of Labor Economics, 19(1):1–20, January 2001.

John T. Addison, McKinley L. Blackburn, and Chad D. Cotti. Do minimum wages raise employment? Evidence from the U.S. retail-trade sector. Labour Economics, 16(4):397– 408, August 2009.

Sylvia A. Allegretto, Arindrajit Dube, and Michael Reich. Do minimum wages really re-duce teen employment? Accounting for heterogeneity and selectivity in state panel data.

Industrial Relations, 50(2):205–240, 2011.

Manuel Arellano and Stephen Bond. Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. The Review of Economic Studies, 58(2):277–297, 1991.

Nathaniel Baum-Snow and Byron F. Lutz. School desegregation, school choice, and changes in residential location patterns by race. The American Economic Review, 101(7):3019– 3046, December 2011.

Pierre Brochu and David A. Green. The impact of minimum wages on labour market tran-sitions. The Economic Journal, 123:1203–1235, December 2013.

Richard V. Burkhauser, Kenneth A. Couch, and David C. Wittenburg. Who minimum wage increases bite: An analysis using monthly data from the SIPP and the CPS. Southern

Economic Journal, 67(1):16–40, July 2000.

Pierre Cahuc and Andreé Zylberberg. Labor Economics. Cambridge: The MIT Press, 2004. David Card and Alan B. Krueger. Minimum wages and employment: A case study of the fast-food industry in New Jersey and Pennsylvania. The American Economic Review, 84 (4):772–793, September 1994.

(25)

Gerald A. Carlino and Richard Voith. Accounting for differences in aggregate state produc-tivity. Regional Science and Urban Economics, 22(4):597–617, November 1992.

Jeffrey Clemens and Michael Wither. The minimum wage and the great recession: Evidence of effects on the wage distributions, employment, earnings, and class mobility of low-skilled workers. Mimeo, September 2014.

Peter A. Diamond. Mobility costs, frictional unemployment, and efficiency. Journal of

Political Economy, 89(4):798–812, August 1981.

Arindrajit Dube, T. William Lester, and Michael Reich. Minimum wage effects across state borders: Estimates using contiguous counties. The Review of Economics and Statistics, 92(4):945–964, 2010.

Arindrajit Dube, T. William Lester, and Michael Reich. Do frictions matter in the labor market? Accessions, separations and minimum wage effects. Working paper 5811, IZA, June 2011.

William E. Even and David A. Macpherson. The wage and employment dynamics of mini-mum wage workers. Southern Economic Journal, 69(3):676–690, January 2003.

Christopher J. Flinn. Minimum wage effects on labor market outcomes under search, match-ing, and endogenous contact rates. Econometrica, 74(4):1013–1062, 2006.

Christopher J. Flinn. The Minimum Wage and Labor Market Outcomes. Cambridge: The MIT Press, 2011.

Laura Giuliano. Minimum wage effects on employment, substitution, and the teenage la-bor supply: Evidence from personnel data. Journal of Labor Economics, 31(1):155–194, January 2013.

Daniel S. Hamermesh. Labor demand and the structure of adjustment costs. The American

Economic Review, 79(4):674–689, September 1989.

Barry T. Hirsch, Bruce E. Kaufman, and Tetyana Zelenska. Minimum wage channels of adjustment. Working paper 6132, IZA, November 2011.

David Lee and Emmanuel Saez. Optimal minimum wage policy in competitive labor markets.

Journal of Public Economics, 96:739–749, 2012.

Jin Young Lee and Gary Solon. The fragility of estimated effects of unilateral divorce laws on divorce rates. Working Paper 16773, National Bureau of Economic Research, February 2011.

Jacob Mincer. Unemployment effects of minimum wages. Journal of Political Economy, 84 (4):87–104, 1976.

(26)

David Neumark and William Wascher. Minimum Wages. Cambridge: The MIT Press, December 2008.

David Neumark, Mark Schweitzer, and William Wascher. Minimum wage effects throughout the wage distribution. The Journal of Human Resources, 39(2):425–450, 2004.

David Neumark, J.M. Ian Salas, and William Wascher. Revisiting the minimum wage-employment debate: Throwing out the baby with the bathwater? Working Paper 18681, National Bureau of Economic Research, January 2013.

Austin Nichols. Causal inference with observational data: Regression discontinuity and related methods in Stata. 2009.

Pia M. Orrenius and Madeline Zavodny. The effect of minimum wages on immigrants’ employment and earnings. Industrial and Labor Relations Review, 61(4):544–563, July 2008.

Marianne E. Page, Joanne Spetz, and Jane Millar. Does the minimum wage affect welfare caseloads? Journal of Policy Analysis and Management, 24(2):273–295, 2005.

David Roodman. How to do xtabond2: An introduction to difference and system GMM in Stata. Stata Journal, 9(1):86–136, 2009.

John Schmitt. Why does the minimum wage have no discernible effect on employment? Working paper, Center for Economic and Policy Research, February 2013.

Isaac Sorkin. Minimum wages and the dynamics of labor demand. Mimeo, February 2013. George J. Stigler. The economics of minimum wage legislation. The American Economic

Review, 36(3):358–365, June 1946.

Gerard J. Van den Berg and Geert Ridder. An empirical equilibrium search model of the labor market. Econometrica, 66(5):1183–1221, September 1998.

Jenny Williams. Does liberalizing cannabis laws increase cannabis use? Journal of Health

Economics, 36:20–32, 2014.

Justin Wolfers. Did unilateral divorce laws raise divorce rates? A reconciliation and new results. The American Economic Review, 96(5):1802–1820, December 2006.

Jeffrey M. Wooldridge. Econometric Analysis of Cross Section and Panel Data. MIT Press, 2002.

Madeline Zavodny. The effect of the minimum wage on employment and hours. Labour

(27)

(a) Treatment effect discrete in levels

(b) Treatment effect discrete in growth

(28)

Figure 2: Frequency of increases to effective state nominal minimum wages (1976-2012)

Figure

Figure 1: Illustration of two types of treatment effects with staggered treatments
Figure 2: Frequency of increases to effective state nominal minimum wages (1976-2012)
Figure 4: Example difference-in-differences without versus with jurisdiction time trends
Table 1: Estimates from Monte Carlo simulation exercise
+7

References

Related documents

The purpose of this binary logistic regression study was to explore the predictive power between an instructor’s employment classification and a student’s gender, race,

An offer to have declined for sure, was the break of combi 3 at 4: this breakout appeared right in the level of a previous high, quite a few pip away from the 25ema, and suffered

Toly Bread (stock code: 603866), as the industry's leading company, is a good example for us to study its operating performance and financial performance, and by doing so, to

There are only a few studies on the causes of media bias and thus this paper uses the data in Chinese Capital Market to empirically study the determinants of media

Utilizing methods including literature review, normative research and multi-case study, this paper takes Vanke, Minsheng Banking and Shanshui Cement Group as examples to

accurately detail ethnicity (see point 3b above), both in relation to offending and staffing requirements; and to resourcing staff training, designed to improve their ability

The large intestine of the human consists of the cecum, ascending colon, transverse colon, descending colon, sigmoid colon, and rectum. Ascending Colon Descending Colon

The Commission will support Member States in promoting a permanent dialogue and peer evaluation at European level on issues such as labour market gaps,